Tejas Dharani

I’m an AI/ML engineer focused on shipping reliable, production-grade LLM systems—especially RAG, agent orchestration, and multi-agent architectures. I care about evaluation, monitoring, and Responsible AI practices so models perform consistently where it matters.

At Forage AI, I architected and deployed end-to-end LLM-powered microservices for enterprise partners, covering ingestion to structured delivery using PyTorch, LangGraph, FastAPI, and AWS/GCP. I built production RAG pipelines enabling sub-10-second queries across tens of thousands of documents, and an Intelligent Document Processing pipeline achieving 97% field-level accuracy with structured output schemas.

I also build for real-world constraints: at Forage AI I reduced LLM operational costs by 70% via token optimization and confidence-based routing, cutting human review queues by 60%. I develop adaptive agentic workflows for reliable web data extraction and support deployments with technical documentation and reference architectures.

Before that, I contributed to the open-source ecosystem and built production AI and backend systems: I helped architect Microsoft AutoGen’s tool-calling loop (reducing agent setup complexity by 60%) and designed streaming multi-agent communication for 10x higher throughput. Earlier roles included building a full-stack AI chatbot platform (FastAPI, React, Docker, CI/CD) and maintaining Python REST APIs for education ERP modules supporting 100+ institutions.